Article Question review

Contents lists available at ScienceDirect

Behavioural Processes

journal homepage: www.elsevier.com/locate/behavproc

Review

On defining resurgence

Kennon A. Lattal⁎,1, Carlos R.X. Cançado2, James E. Cook3, Stephanie L. Kincaid, Tyler D. Nighbor, Anthony C. Oliver West Virginia University, United States

A R T I C L E I N F O

Keywords: Resurgence Recurrent behavior Condition worsening Extinction Definition Operant behavior Reinforcement history Stimulus variables

A B S T R A C T

A review of different investigators’ definitions of resurgence revealed several common features: First, characteristics of the resurgent, or target, response, such as its transience; magnitude; time course within and across sessions; and relativity to a baseline response rate are not mentioned. Second, the target response is described as being established through its reinforcement in the first, or Training, phase of a resurgence procedure. Third, the target response must be eliminated as an alternative response is reinforced in the second, Alternative Reinforcement, phase of a resurgence procedure. Fourth, the alternative response must be extinguished during the Resurgence Test phase. Fifth, none of the definitions allude to any contribution of stimulus variables to resurgence. When reconsidered in light of contemporary research germane to these features, none of the reviewed definitions sufficiently reflect important variables in the generation and assessment of resurgence. The review concludes with a proposed working definition that takes into account contemporary research involving all of the aforementioned factors.

1. The renaissance of resurgence

First systematically investigated by Carey in 1951 and thereafter in the 1980s by Epstein and Skinner (1980) (Epstein, 1983; 1985; see Epstein 2015 for a review of this early research), behavioral resurgence has undergone something of a renaissance beginning in the early years of the 21st century. Much of the interest has been generated in the arenas of translational research and application, where resurgence has been proposed as, among other things, a model for the recurrence of substance abuse following treatment removal (e.g., Craig et al., 2016; Marchant et al., 2013; Podlesnik et al., 2006) and, more generally, of the return of problem behavior on the discontinuation of an ameliora- tive therapeutic program (e.g., Lattal and Peter Pipkin, 2009; Lit and Mace, 2015). Breaking ranks with the classical model of movement from basic research to translation to application (see Neef and Peterson, 2003), like many other behavioral phenomena, since around the turn of the century contemporary basic research on resurgence has proceeded largely in parallel with, rather than preceding, translation and applica- tion. Perhaps this is because once the basic phenomenon was estab- lished its implications were both immediately apparent and compelling. Much of both contemporary research and application related to

resurgence is grounded in earlier conceptualizations, formalized in definitions that are both unnecessarily restrictive and mute with respect to the most recent research on the measurement and controlling variables of resurgence. This review begins with several definitions of resurgence proffered by researchers and practitioners. It then juxta- poses elements of the common features of these definitions with recent research germane to those features, and concludes with a suggested revised working definition.

2. Defining resurgence

Many investigators of the resurgence of operant behavior have begun their analysis with a definition. A representative sample of these definitions is as follows:

− “Resurgence, the reemergence of a previously (but not currently) reinforced response when a subsequently reinforced response is placed on extinction …” (Bruzek et al., 2009, p. 327).

− “Regression: The reappearance of previously extinguished behavior during the extinction of more recently reinforced behavior (sometimes referred to as extinction-induced resurgence)” (Catania, 1991). [Note: Carey (1951) also labeled the phenomenon regression.]

http://dx.doi.org/10.1016/j.beproc.2017.04.018 Received 1 December 2016; Received in revised form 26 April 2017; Accepted 27 April 2017

⁎ Corresponding author at: Department of Psychology, West Virginia University, Morgantown, WV 26506-6040, United States.

1 Now at Universidade de Brasília, Campus Universitário Darcy Ribeiro, 70. 910-900, Brasília-DF, Brazil. 2 Now at University of Mississippi Medical Center, Jackson, MS, 39216 USA. Carlos Cançado was supported by a postdoctoral fellowship (PNPD-CAPES) during the writing of this

article. 3 Now at Department of Psychology, Rollins College, Winter Park, FL 32789 USA.

E-mail address: [email protected] (K.A. Lattal).

Behavioural Processes 141 (2017) 85–91

Available online 06 May 2017 0376-6357/ © 2017 Elsevier B.V. All rights reserved.

MARK

−“Resurgence is the term used here to refer to this occurrence of a previously reinforced behavior … when a more recently reinforced behaviour … is undergoing extinction …” (Cleland et al., 2000, p. 118; see also Cleland et al., 2001).

− “A simple principle of potentially wide application may be stated as follows: When, in a given situation, recently reinforced behavior is no longer reinforced, behaviors that were previously reinforced under similar circumstances tend to recur” (Epstein, 1983).

− “Resurgence is the recovery of a previously reinforced and extinguished response during extinction of a second previously re- inforced response” (Hoffman and Falcomata, 2014).

− “Resurgence occurs when a previously extinguished response recovers following the extinction of some alternative response” (Lambert et al., 2016, p. 283).

− “Resurgence is said to occur when a previously learned response recurs following a hiatus from that response, during which time some other response first is reinforced and thereafter extinguished” (Lattal and St. Peter Pipkin, 2009, p. 254).

− “Resurgence refers to recovery of an extinguished operant response after discontinuation of reinforcement of an alternative response” (Marchant et al., 2013, p. 678).

− “Resurgence is characterized by the reappearance of an extin- guished operant response when an alternative behavior introduced during extinction is subsequently placed on extinction” (Podlesnik and Shahan, 2009).

− “Extinguished operant behavior can return or ‘resurge’ when a response that has replaced it also is extinguished” (Trask et al., 2015, p. 187).

These definitions, along with others, which were omitted simply because they repeat the same points, share several elements. First, measurement characteristics of the resurgent (hereafter, target) re- sponse, such as its magnitude and time course within and across sessions and its relation to some behavioral baseline are not mentioned in any of the definitions. Second, the target response must be established by its reinforcement, which is accomplished in the first, or Training (T), phase of a resurgence procedure. Third, the target response must be at least nominally extinguished as an alternative response is reinforced in the second, Alternative Reinforcement (AR), phase of a resurgence procedure. (In some experiments, this phase is broken into two phases, with functional extinction of the first-trained response occurring in the T phase followed by reinforcement of the alternative response in the second; cf. Cleland et al., 2000; Epstein, 1983.) Fourth, the alternative response must be extinguished during the Resurgence Test (RT) phase. Fifth, none of the definitions allude to the role of stimulus variables in resurgence. The question of how each of these shared definitional features map onto what currently is known about resurgence is considered in the sections that follow.

3. Measuring resurgence

Definitions typically start with measurement (e.g., Bridgman, 1927; Kaplan, 1964/1998; Kaplan, 1964; Plutchik, 1968). Most of the definitions in the preceding section indicate the measurement of discrete responses (as in “a response” or “a behaviour”). A few, however indicate simply “behavior” as the measure. It has been suggested that a good bit of operant behavior, such as reading, conversing, eating, or sexual activity, does not lend itself as readily to discrete frequency counts as it does to time allocated to such activities (e.g., Baum, 2012; Premack, 1965; see also Brownstein and Pliskoff, 1968). But, does operant behavior measured as time allocation resurge similarly to operant behavior measured as discrete responses? Cançado et al. (in press) measured resurgence as time allocated to one of two concur- rently available variable-time reinforcement schedules. In the T phase, one VT schedule delivered response-independent reinforcers with a probability of 0.75 and the other with a probability of 0.25. Each peck on a changeover key shifted the discriminative stimuli and schedules in

effect. The discriminative stimuli associated with the richer and leaner schedules were reversed in the AR phase, with a reversal of the time allocated to the schedules and their associated stimuli. In the RT, reinforcement was eliminated and time allocation transiently shifted in favor of the stimulus associated with the richer schedule in the first phase. Bruzek et al. (2009) reported similar findings with college students. When time spent on activities designed to quiet the crying of a simulated infant no longer quelled its crying, the students spent their time engaged in activities that, in the T phase, had that quieting effect.

The most direct measure of resurgence is an absolute one: How many responses occur or how much time is allocated to an alternative during the RT phase? This widely used index of resurgence can be misleading in the absence of a baseline context. For that reason, most tests of resurgence either informally or formally take into account the rate of the target response relative to its occurrence in other phases of the resurgence experiment. For example, using a single discrete response, the usual procedure is to compare the rate of the target response to its rate of occurrence in the terminal sessions of the immediately preceding AR condition. Resurgence is said to occur only when the target response rate is relatively higher than its rate in those terminal sessions.

Other relative measures of resurgence can be useful when the effects of two or more independent variables or values of independent variables on resurgence are directly compared using either multiple (e.g., Doughty et al., 2007) or concurrent schedules (for further discussion of measuring resurgence, see Cançado et al., 2016). da Silva et al. (2008, Experiment 1b), for example, used a three-phase concurrent resurgence procedure to examine the effects of reinforce- ment frequency of the target response in the T phase on subsequent resurgence. Pigeons first were trained on a concurrent variable-interval (VI) 1-min VI 6-min schedule. In the AR phase, key-peck reinforcement was discontinued (extinction) and reinforcers were delivered according to a single VI 3-min schedule for responding on a third response key. During the RT phase, more responding occurred on the key previously associated with the schedule delivering more frequent reinforcement; however, when responding on the keys formerly associated with the VI 1-min and VI 6-min schedules was normalized with respect to the response rates on either key in the T phase, the relative resurgence on the two keys was about the same.

Two other, less well investigated, dimensions of the measurement of resurgence important in its definition are temporal and local ones. Sidman (1960) observed that “the completion of a transitory phase is marked by a return to the same behavior that would have been observed if the transitory effect had never taken place” (p. 311), and so it is with resurgence. When the RT is extended beyond a single session, the time course of resurgence sometimes, but not always, is one of the target behavior increasing from near zero at the beginning of the RT phase to peak as the alternative response dissipates across this phase. The target behavior then dissipates to zero as the RT phase continues (presumably because such behavior is not reinforced). A common, but not uniform finding, is that resurgence often reaches its apogee not in the first session, but in subsequent ones (depending, of course, on such things as session duration and frequency). Basing assessments of resurgence on single sessions therefore may be mislead- ing in terms of whether it occurred and the extent of such occurrence. Little is known at this point about variables controlling either the time course or qualitative character of resurgence. Session length undoubt- edly has an effect, but whether, and, if so how, these aspects of resurgence change as a function of variables in either the training or alternative reinforcement phases is under-investigated (see Podlesnik and Kelley, 2014; for an assessment of how stimulus variables affect the pattern of resurgence across sessions in the RT phase). In some reports of resurgence, its transitory nature has been ignored by using a single RT-phase session. Even more problematic is that in group designs the highly variable resurgence effect often is averaged across several subjects, obfuscating the variability of its time course.

K.A. Lattal et al. Behavioural Processes 141 (2017) 85–91

86

Even within a session, a report of resurgence as an average of responding over the session duration might potentially mask within- session changes in resurgence. A more molecular analysis of resurgent behavior through cumulative records can provide a clearer picture of the development and denouement of resurgence in real time, both within and across sessions. As an example, the data in Fig. 1, from an unpublished experiment, described below, conducted by Oliver et al. (2017) in which in the RT phase the number of pellets delivered on a VI schedule for the alternative response was reduced from four to one, illustrate the development and time course of resurgent responding during the first three sessions of an RT phase (see also Lieving and Lattal, 2003, Fig. 4). During the first session, resurgence occurred later relative.

to its appearance in the two subsequent sessions. In the second and third sessions, however, the target response occurred earlier in the session but, especially in the third session, dissipated in magnitude rather quickly as each session progressed. Notably absent was any marked burst of responses (i.e., any spontaneous-recovery like effect) at the onset of either the second or third sessions. Qualitative differences in target response occurrence within and across sessions may reveal heretofore uninvestigated properties of resurgence, offering the possi- bility of not only a more nuanced understanding of the variables that affect resurgence and the different ways in which resurgence may be manifest, but also, particularly in applications, other ways of more effectively intervening to attenuate or potentiate resurgence.

4. Transitory resurgence

A critical feature of resurgence absent from all of the definitions cited at the beginning of this review is the transience noted above. This transience presents a methodological challenge in analyzing the con- trolling variables of resurgence, which in turn allow refinement of its definition. Replicating a resurgence effect requires that the different phases of the resurgence procedure be repeated with the same subjects, a process requiring multiple sessions at each phase. Even then, however, subsequent iterations of the procedure may affect the quantity and quality of resurgence (although Lieving and Lattal, 2003, Experi- ment 2, found no systematic differences in resurgence in four pigeons across two direct replications of the three-phase resurgence procedure). Nonetheless, a procedure allowing the repeated generation of resur- gence successively across each of several sessions would obviate the need for this cumbersome, time-consuming across-session repetition of each phase of the resurgence procedure.

Cook and Lattal (2017) developed such a within-session repeated resurgence procedure by dividing each of 30 successive individual sessions into three-phase resurgence procedures. In the T phase, target- key pecking was reinforced on a fixed-interval (FI) 30-s schedule until at least 10 reinforcers were collected and responding on two other

response. Keys, one labeled the alternative response, and the other a sham

(i.e., control) key on which responses were never reinforced (see below) was near zero. In the AR phase, reinforcement of the target response was discontinued, and responses on a second, alternative, key were reinforced according to an FI 30-s schedule until at least 10 reinforcers were collected and until responding on the FI key and the sham key were near zero. In the RT phase, reinforcement of the alternative response was discontinued. The RT phase continued until responding ceased on all three response keys, at which point the session termi- nated. The data in Fig. 2 show the results across successive sessions for one of the four pigeons used in the experiment (each with results similar to the one shown here). Target-key responding reliably in- creased (resurgence), relative to its occurrence in the AR phase, during the RT phase during most sessions. Of note is the gradually declining magnitude (represented by the height of the bars) and occurrence (represented by the number of bars) of the resurgence effect across sessions. A decline in magnitude within sessions also occurred, but responding often peaked in the second or later block, rather than the first RT-phase block. This procedure produced resurgence reliably both within and across sessions, but what factors account for the occurrence of resurgence in some but not all sessions and the change in the resurgence effect across replications remains a question for future investigation.

Repeated within-session resurgence tests might be developed with any procedure where repeated exposure to extinction occurs within individual sessions and the extinction effect can be repeated reliably within and across successive sessions. In a progressive-ratio (PR) schedule, for example, each session remains in effect until extinction occurs (i.e., a break point of a predetermined time period without a response). Using a variation of the Cook and Lattal procedure outlined above, within a single session a target response might be reinforced according to some schedule of reinforcement in the T phase. This could be followed by extinction of the target response and reinforcement of an alternative response according to a progressive-ratio schedule, the response requirement of which would increment until a break point is reached, with resurgence occurring at some point during the ratio progression (see also Jarmolowicz and Lattal, 2014). As another example, Bai et al. (2017) adapted the free-operant psychophysical procedure to study resurgence within-sessions by embedding probe

Fig. 1. Cumulative target responses of a rat during the first three sessions of an RT phase. Each hash mark on the x-axis represents 100 s.

Fig. 2. Absolute response rates across successive 5-interval blocks on the target key in a within-session repeated resurgence experiment conducted by Cook and Lattal. Individual sessions are shown on the z-axis with more recent sessions towards the front. The leftmost bar in each session is the absolute response rate during the last 5-interval block of the AR Phase. The remaining bars are successive 5-interval blocks of the RT Phase.

K.A. Lattal et al. Behavioural Processes 141 (2017) 85–91

87

trials consisting of extended local periods of extinction among trials that typically reinforced a target followed by an alternative response. Restricting the definition of resurgence to transient observations of its appearance and then disappearance provides only a narrow picture of the phenomenon, and hence its definition. Expanding the resurgence paradigm by attempting to create “steadier states” of resurgence opens possibilities not only for expanding the definition, but also its implica- tions of resurgence for reinforcement theory and practice.

5. Target-response reinforcement history as a defining feature of resurgence

In each of the definitions cited in the Defining resurgence section above, the target response is described as having been established in the organism’s repertoire at some previous time, although the conditions defining “established” and “some previous time” are not specified. The relation between the conditions of establishment of the target response and its subsequent resurgence is largely unexplored (but cf. Lieving and Lattal, 2003, Experiment 1). Nonetheless, the broad requirement that the target response must have been established in the past through reinforcement precludes including simply any increase in behavior, as in, for example, an increase in general activity, in the definition of resurgence. Also excluded from these definitions are increases in specific responses that have not previously been reinforced. Stated another way, a response has to have a history of reinforcement before it can be subject to resurgence.

One way of affirming that the previous reinforcement history of a specific target response is critical in the later appearance of resurgence has been to include a control or sham operandum throughout the experiment on which responding is never reinforced (see Epstein, 1983; for an example). Responding on the sham operandum during the resurgence test is minimal relative to that observed on the target operandum, which is taken as evidence of the necessity of prior training of the target response. Both a strength and a limitation of this interpretation, however, is that, unlike the target response, there is no history of reinforcement of responding on the sham operandum. Does, then, the absence of responding on this sham operandum reflect the absence of the history of responding associated with the resurgence operandum? Or, in lay terms, does it reflect a simple failure of the operandum as a salient feature of the environment, no different, for example, than many other static features of the environment bearing no particular relation to reinforcement? da Silva et al. (2008, Experiments 1a & 1b) addressed this question using the concurrent resurgence procedure noted above. In Experiment 1a pecking on either of two side keys was autoshaped (Brown and Jenkins, 1968) to provide a minimal history of responding on them. In the next, AR, phase responding on a third key was reinforced according to a VI 1-min schedule. In the subsequent RT, responding on the center key was extinguished and virtually no pecks occurred on either side key, suggesting that any general increase in activity induced by the extinction of responding on the third key was not manifest as respond- ing on the side keys. In Experiment 1b, using the same pigeons, a conventional three-phase resurgence procedure was effected as de- scribed in the Measuring resurgence section above, with the result that now resurgence occurred on both side keys in the RT phase. These results in combination with those of their Experiment 1a suggest the importance of a prior history of reinforcement for the resurgence of responding.

6. Extinction of the target response as a defining feature of resurgence

Although a response has to be established before it can be eliminated, is it necessary that the target response be extinguished before it can be resurged, as each of the definitions in the Defining resurgence section state? Mechner and Jones (2015) omitted this facet of

resurgence in their definition of resurgence as the “reappearance of behavior that occurred earlier in the individual’s history but not recently, without restoration of the conditions under which the earlier behavior occurred” (p. 63). Their definition is unique relative to the others cited previously in that it does not specify extinction, elimina- tion, or suppression of the target response, but rather only that it not be present (reflected by the term “reappearance,” which also implies its current absence) at the time of the resurgence test. (In Mechner & Jones’s definition, “not recently” remains undefined, too.)

In many, perhaps most, instances the “extinction” of the target response in the AR phase is procedural in that reinforcement of that response is eliminated, but it is functional only to the extent that the rate of the target response is near-zero. Whether the response actually is eliminated remains a subject of debate (e.g., Leitenberg et al., 1975). There are two procedures used to study resurgence, in which the target response in the AR phase may be considered “eliminated” and not just nominally so. One is when the target response is reduced to near zero before the alternative response is reinforced, as in the four-phase resurgence procedure described above (cf. Epstein, 1983; Lieving and Lattal, 2003). The other is when a DRO schedule is used in the AR phase, as was done by da Silva et al. (2008, Experiments 2 & 3). Under a DRO schedule, the target response cannot be reinforced, thus the target response must be functionally eliminated to the point that a pause in responding is sufficiently long that the DRO-scheduled reinforcer occurs. Either of these two procedures to explicitly eliminate the target response results in its subsequent recurrence in the RT phase. However, research comparing resurgence generated using either of these proce- dures may yield resurgence differing qualitatively and quantitatively from one another (cf. Doughty et al., 2007).

7. Extinction of the alternative response as a defining feature of resurgence

All of the definitions in the section on Defining resurgence note that resurgence occurs when the alternative response is extinguished. But is such extinction really necessary for generating resurgence? The answer suggested by several recent experiments is “no.” Lieving and Lattal (2003, Experiment 4) reported resurgence when, in the resurgence test, rather than extinguishing the alternative response, the reinforcement schedule associate with responding in the alternative reinforcement condition was changed from the VI 30-s schedule that was in effect in the AR phase to a VI 6-min schedule in the RT phase. The resurgence effect under the latter condition occurred in 2 of 3 pigeons, and the effect was weaker than that obtained when, in the RT phase, the alternative response was extinguished. At least some resurgence has been found in our laboratory when the conditions of reinforcement of the alternative response are “worsened” in other ways than reductions in reinforcement rate. Resurgence occurred reliably, for example, when the number of food pellets delivered to rats following lever-press responses was reduced from four in the AR phase to one in the RT phase (see also Craig et al., 2017, Experiment 2). Results of the first RT session for one rat in this experiment are shown in Fig. 3.

In a related experiment with pigeons, however, when the duration of access to a food reinforcer was reduced from 6 s in the AR phase to 1 s in the RT phase, resurgence occurred in only four of seven pigeons, suggesting that magnitude reductions as an inducer of resurgence may depend on the method by which the magnitude is reduced.

Nighbor et al. (2017b) investigated how delayed reinforcement affects resurgence (see also Jarmolowicz and Lattal, 2014). Changing the immediate reinforcement of the alternative response in the AR phase to reinforcement delivered after a 60-s (blackout) signaled delay in the RT phase yielded resurgence of the target response in each of six pigeons exposed to this procedure. Reinforcement rates between the immediate and delayed reinforcement conditions differed by less than one reinforcer per minute. The important finding for the present discussion, however, is that, as in Lieving and Lattal’s Experiment 3

K.A. Lattal et al. Behavioural Processes 141 (2017) 85–91

88

and the reduction in the number of food pellets delivered to rats from the AR to the RT phase, in this experiment reinforcement conditions worsened, either by adding delays or decreasing reinforcement rate, and the result was resurgence.

All of the findings discussed thus far in this section suggest that resurgence results when the reinforcement conditions are worsened during the transition from the AR to RT phases, that is, by introducing extinction, reducing reinforcement rate, decreasing reinforcer magni- tude, or adding delays to reinforcement. Oliver et al. (2017) examined whether resurgence also might occur when reinforcement rate is not reduced between the AR and RT phases. Following establishment of a target keypeck response, that response was extinguished in the AR phase as a keypeck to an alternative operandum was reinforced according to a VI schedule. In the RT phase, extinction remained in effect for the target response but, instead of extinguishing the alter- native response, when reinforcement became available under the VI schedule, a unique stimulus (keylight color) appeared and remained on until a response produced the reinforcer. Under these conditions, alternative-response rates dropped precipitously, but the rate of reinforcement remained unchanged from the AR phase. Despite the constant reinforcement rate between the AR and RT phases, resurgence occurred in each of the four pigeons exposed to the procedure. Although technically there was no worsening of the reinforcement conditions between the AR and RT phases, it could be argued that the signal rendered the remainder of the VI schedule and its associated stimulus a period of extinction. These results nonetheless stand in contrast to the assertion in most definitions of resurgence that removing reinforcement for the alternative response is necessary for its occur- rence.

Taken together, the fact that reductions in reinforcement frequency and magnitude, and increases in delays all yield at least some resurgence suggest that definitions attributing resurgence to the extinction of the alternative response are unnecessarily narrow and fail to capture the breadth of the resurgence effect. Rather, it appears, as Lattal and Wacker (2015) earlier suggested (see also Epstein, 1985; Cleland et al., 2001) that resurgence can result whenever the alter- native conditions of reinforcement are worsened in the RT phase relative to those in the AR phase.

Two other issues germane to the worsening of conditions as a precipitator of resurgence also warrant comment. One is defining “worsening,” which, admittedly is a vague term. A reasonable candi- date for defining “worsening” is any reinforcement condition not

preferred relative to its alternative in a choice situation. The other issue is that response rates generally vary directly with reinforcement rate and reinforcement magnitude and inversely with delay. Thus, manipulating these reinforcement parameters also changes response rates. In one experiment, da Silva et al. (2008, Experiment 3) arranged concurrent reinforcement of key pecking of pigeons under higher or lower reinforcement rates. Schedule contingencies were such that response rate on the two keys was approximately equal despite the different reinforcement rates. Resurgence in the subsequent RT phase was similar between keys, suggesting that resurgence is “better predicted by the rate at which responding previously occurred … than [by] the rate at which … responding was reinforced (da Silva et al., 2008). Their findings in combination with the research reviewed in this section suggest that the relation between response rate per se and resurgence merits further experimental analysis.

8. Stimulus conditions and resurgence

None of the resurgence definitions reviewed above mention the role of stimuli in its occurrence. Stimulus control nonetheless is involved in resurgence (e.g., King and Hayes, 2016; Podlesnik and Kelley, 2014). Bouton and colleagues (e.g., Winterbauer and Bouton, 2010; Bouton and Trask, 2015) have suggested that the conditions in effect in the different phases of the resurgence procedure might be construed as different contexts (i.e., stimuli) evoking the resurgent response in the RT phase of a resurgence procedure. Although the contextual-change theory of recurrence is parsimonious and broadly encompassing, it has its fair share of criticisms (Lattal and Wacker, 2015; McConnell and Miller, 2014). Lattal and Wacker (2015), for example, observed that

…if one institutes a nominal ABC renewal procedure, but fails to obtain renewal in the C condition, does one conclude that such renewal does not occur or that the C condition did not really constitute a context change? If the latter, then a C condition exists only if renewal occurs, making renewal both the definition and the cause of the recurrence. (p. 4).

Aside from these more general considerations, the ‘arranged exter- oceptive’ stimuli in most experiments on resurgence remain fixed during each of the phases, which may account for the absence of allusion to stimulus conditions in the definitions cited above. Thus, for example, if a red keylight is associated with the alternative operandum in the T phase, that same red keylight remains in effect on this operandum in both the AR and RT phases. Using a concurrent resurgence procedure (cf. da Silva et al., 2008; Experiments 2 and 3), Kincaid et al. (2015) transilluminated both response keys white in the T phase. During the AR phase, with DRO schedules in effect on both response keys, one of the keylights remained white and the other was changed to red. In the RT phase, both keylights again were white. Using this procedure it was possible to compare directly resurgence under conventional stimulus conditions (white lights in each phase were the same, what might be labeled AAA) to that produced by an ABA renewal procedure (white-red-white in T, AR, and RT phases, respectively). Resurgence occurred under both stimulus conditions; however, the added ABA renewal procedure produced far more resurgence (i.e., far more responses in extinction) than did the AAA conventional stimulus condition resurgence procedure. Using the same concurrent resurgence procedure as Kincaid et al., during the RT phase, Nighbor et al. (2017a) associated one of the alternatives with a novel stimulus in both the AR and RT phases (creating an ABC resurgence test) while the stimulus associated with the other response key remained as it had been during the T and AR phases (an AAA resurgence test). Fig. 4 shows response rates per minute on both the AAA and ABC target response keys averaged across the last six AR phase and across the last six (Pigeons 3181 and 9553) or five (Pigeon 2748) RT sessions. Each pigeon responded more frequently during the RT phase on the response key associated with AAA than on the concurrently available one associated with ABC resurgence, again suggesting the importance of stimulus

Fig. 3. Cumulative lever press responses on the vertical and horizontal levers across successive seconds of the first session of the RT phase. At the start of this session, the number of pellets per reinforcer for responding on the vertical lever was reduced from 4 pellets to 1. Responses to the horizontal lever, the target response, were not reinforced. Vertical dotted lines indicate reinforcers delivered following responses on the vertical lever. The gradual development of resurgence across the session is shown, as is a tendency for bursts of horizontal lever presses to occur following reinforcement of vertical lever presses. Each hash mark on the x-axis represents 100 s.

K.A. Lattal et al. Behavioural Processes 141 (2017) 85–91

89

conditions in resurgence. A final observation on this latter topic comes from the experiment

previously described in the section on Extinction of the alternative response as a defining feature of resurgence in which signaling reinforcer availability for the alternative response resulted in resurgence of the target response. As noted, from the standpoint of reinforcement availability, conditions did not worsen relative to the previous condi- tion, but simply adding a stimulus correlated with nonreinforcement was sufficient to evoke resurgence.

9. Conclusion: refining the definition

The definition of any behavioral phenomenon must navigate between precision and generality. It must be consistent with extant data, but at the same time cannot be so precise as to constrain the phenomenon within narrow limits of only what is known. Common features of the resurgence definitions reviewed in the Definitions of resurgence section were the topographical and temporal characteristics of the resurgent behavior, establishment of and subsequent frequency reduction of the target response, extinction of the alternative response as the occasion for resurgence, and the role of stimulus variables in resurgence. Some of those definitions go back to the early days of resurgence research and thus are unsurprisingly out of tune with more contemporary findings. Considering these features in light of the current understanding of the controlling variables of resurgence suggests a definition along the following lines: Resurgence currently can be understood as the transient recurrence, with consideration of the stimulus context, of some dimension of previously established but not currently occurring activity when reinforcement conditions of current behavior are worsened. This definition exorcises resurgence from the metaphorical death grip that extinction has held over it by not mentioning either why the reappearing behavior is not presently there or limiting the conditions that might evoke its reappearance to the absence of reinforcement of the alternative response. The definition also captures the transience of resurgence and alludes to “activity” rather than “response,” allowing for the possibility of operants not defined as discrete responses as well as other measures of resurgence such as latency or response topography variation. And, lastly, but

importantly, it allows a role for stimulus variables in resurgence. Definitions of behavioral phenomena are always works in progress

because research on those phenomena always is a work in progress. Indeed, the goal of definition is not to achieve the impossible of developing the definition of something. Rather it is to articulate a useful definition consistent with what is known, which both stimulates further experimentation and also allows the results to coalesce around a common framework, while at the same time being sufficiently circum- spect to allow the understanding of the phenomenon to emerge from an experimental analysis of its controlling variables instead of a priori pronouncements about the nature and limits of those variables. It is in this spirit that we offer the present review and the definition that emerges from it.

References

Bai, J.Y.H., Cowie, S., Podlesnik, C.A., 2017. Quantitative analysis of local-level resurgence. Learn. Behav. 45, 76–88.

Baum, W.M., 2012. Rethinking reinforcement: allocation, induction: and contingency. J. Exp. Anal. Behav. 97, 101–124.

Bouton, M.E., Trask, S., 2015. Role of the discriminative properties of the reinforcer in resurgence. Learn. Behav. 44, 137–150.

Bridgman, P.W., 1927. The Logic of Modern Physics. MacMillan, New York. Brownstein, A.J., Pliskoff, S.S., 1968. Some effects of relative reinforcement rate and

changeover delay in response-independent concurrent schedules of reinforcement. J. Exp. Anal. Behav. 11, 683–688.

Bruzek, J.L., Thompson, R.H., Peters, L.C., 2009. Resurgence of infant caregiving responses. J. Exp. Anal. Behav. 92, 327–343.

Cançado, C.R.X., Abreu-Rodrigues, J., Aló, R.M., 2016. A note on measuring recurrence. Mexican J. Behav. Anal. 42, 75–86.

Cançado, C.R.X., Lattal, K.A., Carpenter, H.K., Solley, E.A., 2017. Resurgence of time allocation. J. Exp. Anal. Behav (in press).

Carey, J.P., 1951. Reinstatement of previously learned responses under conditions of extinction: a study of regression (Abstract). Am. Psychol. 6, 234.

Catania, A.C., 1991. Glossary. In: Iverson, I., Lattal, K.A. (Eds.), Experimental Analysis of Behavior: Part 2. Elsevier, Amsterdam, pp. G1–G44.

Cleland, B.S., Foster, T.M., Temple, W., 2000. Resurgence: the role of extinction. Behav. Process. 52, 117–129.

Cleland, B.S., Guerin, B., Foster, T.M., Temple, W., 2001. On terms: resurgence. Behav. Anal. 24, 255–260.

Cook, J.C., Lattal, K.A., 2017. Repeated Within-Session Resurgence. (Manuscript submitted for publication).

Craig, A.R., Nall, R.W., Madden, G.J., Shahan, T.A., 2016. Higher rate alternative non- drug reinforcement produces faster suppression of cocaine seeking but more resurgence when removed. Behav. Brain Res. 306, 48–51.

Craig, A.R., Browning, K.O., Nall, R.W., Marshall, C.M., Shahan, T.A., 2017. Resurgence and alternative reinforcement magnitude. J. Exp. Anal. Behav. 107, 218–233.

Doughty, A.H., da Silva, S.P., Lattal, K.A., 2007. Differential resurgence and response elimination. Behav. Process. 75, 115–128.

Epstein, R., Skinner, B.F., 1980. Resurgence of responding after cessation of response- independent reinforcement. Proc. Natl. Acad. Sci. U.S.A. 77, 6251–6253.

Epstein, R., 1983. Resurgence of previously reinforced behavior during extinction. Behav. Anal. Lett. 3, 391–397.

Epstein, R., 1985. Extinction-induced resurgence: preliminary investigations and possible applications. Psychol. Record 35, 143–153.

Epstein, R., 2015. On the rediscovery of the principle of resurgence. Mexican J. Behav. Anal. 41, 19–43.

Hoffman, K., Falcomata, T.S., 2014. An evaluation of resurgence of appropriate communication in individuals with autism who exhibit severe problem behavior. J. Appl. Behav. Anal. 47, 651–656.

Jarmolowicz, D.P., Lattal, K.A., 2014. Resurgence under delayed reinforcement. Psychol. Record 64, 189–193.

Kaplan, A. (1964/1998). The conduct of inquiry. New Brunswick, NJ: Transaction Publishers

Kincaid, S.L., Lattal, K.A., Spence, J., 2015. Super-resurgence: combining resurgence with renewal. Behav. Process. 115, 70–73.

King, J.E., Hayes, L.J., 2016. The role of discriminative stimuli on response patterns in resurgence. Psychol. Record 66, 325–335.

Lambert, J.M., Bloom, S.E., Samaha, A.L., Dayton, E., Kunnavatana, S.S., 2016. Effects of noncontingent reinforcement on the persistence and resurgence of mild aggression. Psychol. Record 66, 283–289.

Lattal, K.A., Peter Pipkin St., C., 2009. Resurgence of previously reinforced responding: research and application. Behav. Anal. Today 10. http://www.behavior-analyst- today.net.

Lattal, K.A., Wacker, D., 2015. Some dimensions of recurrent operant behavior. Mexican J. Behav. Anal. 41, 1–13.

Leitenberg, H., Rawson, R.A., Mulick, J.A., 1975. Extinction and reinforcement of alternative behavior. J. Comp. Physiol. Psychol. 88, 640–652.

Lieving, G., Lattal, K.A., 2003. Recency, repeatability: and reinforcement retrenchment: an experimental analysis of resurgence. J. Exp. Anal. Behav. 80, 217–233.

Lit, K., Mace, C., 2015. An example of translational research on recurrence of operant

Fig. 4. Mean responses per minute on either target response key for each of three pigeons during the AR and RT phases of a concurrent ABC AAA resurgence procedure.

K.A. Lattal et al. Behavioural Processes 141 (2017) 85–91

90

behavior and treatment relapse. Mexican J. Behav. Anal. 41, 269–288. Marchant, N.J., Li, X., Shaham, Y., 2013. Recent developments in animal models of drug

relapse. Curr. Opin. Neurobiol. 23, 678–683. McConnell, B.L., Miller, R.R., 2014. Associative accounts of recovery-from-extinction

effects. Learn. Motiv. 46, 1–15. Mechner, F., Jones, L.D., 2015. Effects of repetition frequency on operant strength and

resurgence of non-criterial features of operants. Mexican J. Behav. Anal. 41, 63–83. Neef, N., Peterson, S.M., 2003. Developmental disabilities: scientific inquiry and

interactions in behavior analysis. In: Lattal, K.A., Chase, P.N. (Eds.), Behavior Theory and Philosophy. Kluwer Academic/Plenum, New York, pp. 369–389.

Nighbor, T.D., Kincaid, S.L., O ’Hearn, C.M., Lattal, K.A., 2017a. Combinations of Resurgence and Two Types of Renewal Procedures. Department of Psychology, West Virginia University, Morgantown WV, USA (Unpublished manuscript).

Nighbor, T.D., Oliver, A.C., Lattal, K.A., 2017b. Delay of Reinforcement Generates Resurgence. Department of Psychology, West Virginia University, Morgantown WV, USA (Unpublished manuscript).

Oliver, A.C., Nighbor, T.D., Lattal, K.A., 2017. Resurgence Without a Change in Reinforcement Rate. Department of Psychology, West Virginia University,

Morgantown WV, USA (Unpublished manuscript). Plutchik, R., 1968. Foundations of Experimental Research. Harper & Row, New York. Podlesnik, C.A., Kelley, M., 2014. Resurgence: response competition, stimulus control:

and reinforcer control. J. Exp. Anal. Behav. 102, 231–240. Podlesnik, C.A., Shahan, T.A., 2009. Behavioral momentum and relapse of extinguished

operant behavior. Learn. Behav. 37, 357–364. Podlesnik, C.A., Jimenez-Gomez, C., Shahan, T.A., 2006. Resurgence of alcohol seeking

produced by discontinuing non-drug reinforcement as an animal model of drug relapse. Behav. Pharmacol. 17, 369–374.

Premack, D., 1965. Reinforcement theory. In: Levine, D. (Ed.), Nebraska Symposium N Motivation. University of Nebraska Press, Lincoln, pp. 123–188.

Sidman, M., 1960. Tactics of Scientific Research. Basic Books, New York. Trask, S., Schepers, S.T., Bouton, M.E., 2015. Context change explains resurgence after

the extinction of operant behavior. Mexican J. Behav. Anal. 41, 187–210. Winterbauer, N.E., Bouton, M.E., 2010. Mechanisms of resurgence of an extinguished

instrumental behavior. J. Exp. Psychol. Anim. Behav. Process. 36, 343–353. da Silva, S.P., Maxwell, M.E., Lattal, K.A., 2008. Concurrent resurgence and remote

behavioral history. J. Exp. Anal. Behav. 90, 313–331.

K.A. Lattal et al. Behavioural Processes 141 (2017) 85–91

91